IntroductionCourse logisticsMalthusian …IntroductionCourse logisticsMalthusian modelMoriscosPapers fro the essays Grading and Requirements Essays (50%) wTo essays about paper-pairs

Lecture 1Introduction

Matti Sarvimäki

History of Economic Growth and Crisis25 February 2014

Introduction Course logistics Malthusian model Moriscos Papers for the essays

Why history?

�[Ancient Greeks] saw the future as something that came uponthem from behind their backs with the past receding away beforetheir eyes. When you think about it, that's a more accuratemetaphor than our present one. Who really can face the future? Allyou can do is project from the past, even when the past shows thatsuch projections are often wrong� (Robert M. Pirsig: Zen and the Art ofMotorcycle Maintenance, 1974, afterword)

�Our basic message is simple: We have been here before. [...]Recognizing these analogies and precedents is an essential steptoward improving our global �nancial system, both to reduce therisk of future crisis and to better handle catastrophes when theyhappen.� (Reinhart & Rogo�: This Time Is Di�erent: Eight Centuries of FinancialFolly, 2009, Introduction)

Matti Sarvimäki Economic History 1: Introduction 1 / 35


Why history?

A design of experiments [...] is an essential appendix to any quantitativetheory. [Experiments] may be grouped into two di�erent classes, namely,(1) experiments that we should like to make to see if certain realeconomic phenomena�when arti�cially isolated from "otherin�uences"�would verify certain hypotheses, and (2) the stream ofexperiments that Nature is steadily turning out from her own enormouslaboratory, and which we merely watch as passive observers. (Haavelmo:The Probability Approach in Econometrics, 1944, Econometrica, 12:1�115)


The Question: How did we become so rich?

An anecdote illustrating how much better o� we are than people just 150 years ago:Prince Albert, the husband of Queen Victoria, died of typhoid at the age of 42 in1861. It is a a bacterial disease transmitted by contaminated food or water (typicallydrinking water being polluted by sewage) and was common in the 19th century cities.Typhoid has virtually disappeard from rich countries due to vaccinations and betterpublic sanitation such as chlorination of drinking water.


What is to be explained?Mortality under age one per 1,000 children in Finland (Statistics Finland)

0

50

100

150

200

250

300

350

400

1750 1800 1850 1900 1950 2000


What is to be explained?Real wages of English building workers (Clark 2005, JPE)

1308 journal of political economy

Fig. 1.—Builders’ real day wages, 1209–2004 (source: table A2)

wage observations and 110,000 observations on prices and housingrents—the details are given in the Appendix. The body of the paperconcentrates on the resultant series and their implications.

The most important implications are that the break from the Mal-thusian era of little advance in efficiency in England began circa 1640,long before the famous Industrial Revolution, and before even the emer-gence of the modern political regime in England in 1689. Further, whileit is possible that the fundamental cause of this break was much greaterinvestment in human capital, those gains in human capital investmentcannot have their origin in the incentives provided by labor markets.Both real wages and the premium for skills in the seventeenth centurydid not change in such a way as to induce a switch to fewer childrenof higher quality. Finally, the new series suggest that the classic IndustrialRevolution of the eighteenth century was much more favorable to work-ers’ real earnings than other recent studies have implied.

Figure 1 shows the estimated real purchasing power of the hourlywage of building workers from 1209 to 2004 by decade. Before 1870,when wages are mainly quoted by the day, work hours are assumed tobe 10 per day. The Appendix shows that, if anything, work hours before1800 were possibly higher than 10 per day, so that the gain in real wagesin the Industrial Revolution era is perhaps greater than the figuresuggests.

Before 1800, though there were major fluctuations, real wages displayno clear trend. Wages in 1200–1249, for example, averaged only 9 per-

This content downloaded from 130.233.243.231 on Mon, 29 Jul 2013 15:10:53 PMAll use subject to JSTOR Terms and Conditions

Based on 46,000 quotes of day wages, 90,000 quotes of the prices of 49 commodities,and 20,000 quotes of housing rents


What is to be explained?GDP per Capita 1�2008 (www.ggdc.net/maddison)

0

5

10

15

20

25

30

GD

P pe

r ca

pita

(th

ousa

nds

of 1

990$

)

0 200 400 600 800 1000 1200 1400 1600 1800 2000

Western Europe Western Offshotts Asia Africa



What is to be explained?GDP per capita in Finland, Sweden and the U.S. (www.ggdc.net/maddison)

Ghana (2008)

Morocco (2008)

Hungary (2008)

1917 1939 1970

1

2

4

8

16

32

GD

P pe

r ca

pita

(th

ousa

nds

of 1

990$

)

1820 1840 1860 1880 1900 1920 1940 1960 1980 2000

Finland Sweden United States



Explaining Economic GrowthMacroeconomics I: Growth and Cycles

Aggregate production is

Y = Af (K , L,X )

where A is total factor productivity and f (·) is a constant returnsto scale production function (factors: capital, labor, land).

Standard growth models provide a valuable framework to thinksystematically about the proximate causes of growth

This course complements Macro 1 by asking:if technology, physical capital and human capital are soimportant, why do they di�er across countries?



Explaining Economic GrowthThis Course

�...the factors we have listed (innovation, economies of scale,education, capital accumulation etc.) are not causes of growth;they are growth.� (North and Thomas: The Rise of the Western World: A NewEconomic History, 1973, p. 2)

�There must be some other reasons underneath those, reasonswhich we will refer to as fundamental causes of economic growth.It is these reasons that are preventing many countries frominvesting enough in technology, physical capital and humancapital.� (Acemoglu: Introduction to Modern Economic Growth, 2007, Ch. 4)



Outline of the course

1 The Malthusian Era (today)

2 Fundamental causes of growth

1 Luck2 Geography3 Culture4 Institutions

3 Innovation and crises

1 Technology2 Finance

4 Unleashing talent

1 Migration2 Intergenerational mobility3 Marriage, family and work



We will use four complementary approaches

1 Narrative history

the �story� of what happened and why

2 Formal models

internally consistent simpli�cations of the complex reality

3 Descriptive empirical work

careful measurement of what actually happened

4 Causal empirical work

(natural) experiments and/or quantitative economic modelsused to compare counterfactual states of the world



Grading and Requirements

Essays (50%)

Two essays about paper-pairs listed in the syllabusMax. 2,000 words (3,000 for PhD students), need to answer:

What are the �take-aways�?What are the key assumptions? Are they plausible (and why)?How do the papers contradict/complement each other?

You are encouraged to work in teams of up to three persons(two for PhD students)The essays are due to March 13th and April 10th

Final exam (50%)

You are expected to know the material covered in the lecturesLecture notes at: www.aalto-econ.�/sarvimaki/courses.html



The Malthusian model

In order to understand why some countries started to growrapidly about two centuries ago, we �rst need to understandwhy they failed to do so earlier

The main explanation is due to Reverend Thomas R. Malthus,whose book An Essay on the Principle of Population was �rstpublished in 1798.



The Malthusian model

Malthus' model builds on the key assumptions that:

population increases with per capita income�xed supply of land leads to negative returns to scaleslow technological progress (implicit)

Given these assumptions, the only way to achieve a permanentincrease in per capita income is to restrict population

e.g. technological innovation increases productivity → wagesincrease → population grows → land/labor ratio decreases →wages decrease back to subsistance level


Steady-State in the Standard Malthusian ModelVoigtländer and Voth (2013, REStud)

[14:26 4/4/2013 rds034_online.tex] RESTUD: The Review of Economic Studies Page: 776 774–811

776 REVIEW OF ECONOMIC STUDIES

Wage

Steady state in the standard Malthusian model Steady states with "Horsemen effect"

Dea

th/B

irth

rate birth rate

death rate

C

wC Wage

Dea

th/B

irth

rate

wH

w0

death rate

birth rate

EU

E0

EH

Figure 1

Steady states in the standard Malthusian model and with “Horsemen effect”

it converges to EH . This is equivalent to a “ratchet effect”: Wage gains after the Black Deathbecame permanent.6

The crucial feature to obtain multiple steady states in Figure 1 is an upward-sloping partof the death schedule. This reflects the historical realities of early modern Europe.7 Accordingto Malthus (1826), factors reducing population pressure include “vicious customs with respectto women, great cities, unwholesome manufactures, luxury, pestilence, and war.” We focus onthree—great cities, pestilence, and war. All of them increased in importance after the plaguebecause of higher per capita incomes. High wages were partly spent on manufactured goods,mainly produced in urban areas. Cities in early modern Europe were death-traps, with mortalityfar exceeding fertility rates. Thus, new demand for manufactures pushed up aggregate death rates.War and trade reinforced this effect. Between 1500 and 1800, the continent’s great powers werefighting each other on average for nine years out of every ten (Tilly, 1992). This was deadly mainlybecause armies on the march often spread epidemics. Wars could be financed more easily whenper capita incomes were high. The difference between income and subsistence increased, leavingmore surplus that could be spent on war. In effect, war was a “luxury good” for princes. In addition,trade grew as people became richer, and it also spread germs.8 In this way, the initial rise in wagesafter the Black Death was made permanent by the ‘Horsemen effect,’ pushing up mortality ratesand producing higher per capita incomes. Thus, Europe experienced a simultaneous rise in warfrequency, in deadly disease outbreaks, and in urbanization. The Horsemen of the Apocalypseeffectively acted as “Horsemen of Riches”.9

6. A large positive shock to technology could theoretically also cause this transition. However, pre-modern ratesof productivity growth are much too low to trigger convergence to the high-income steady state.

7. The theoretical and historical conditions under which the “Horsemen effect” led to multiple steady states arediscussed in detail in Section 2.

8. Numerous studies have focused on the interaction between domestic armed conflict and income. Many findthat civil wars decline in frequency after positive growth shocks (Collier and Hoeffler, 1998, 2004; Miguel et al., 2004).In contrast, Grossman (1991) has argued that higher incomes should promote wars (“rapacity” effect), as there is moreto fight over. Martin et al. (2008) find that more multilateral trade can lead to more war.

9. This is the opposite of the negative effect of wars, civil wars, disease, and epidemics on income levels foundin many economies today (cf. Murdoch and Sandler, 2002; Hoeffler and Reynal-Querol, 2003). The main reason for thisdifference is that human capital is crucial for development today, while it was not in pre-modern times, when decreasingreturns to (unskilled) labour in agriculture dominated the production pattern.

by guest on July 29, 2013http://restud.oxfordjournals.org/

Dow

nloaded from

�Death rates are downward sloping in income, and birth rates are either �at orupward sloping. This generates a unique steady state (C) that pins down wagesand population size. Decreasing marginal returns to labour set in quickly aspopulation grows because �xed land is an important factor of production. Adecline in population can raise wages, moving the economy to the right of C.However, the increase in output per capita is only temporary. Birth rates nowexceed death rates and population grows, which in turn will depress wages�the�Iron Law of Wages� holds.�


The Malthusian model: Discussion

Malthus presented the �rst coherent theory about how limitedresources constraint economic growth

it is often mistakenly thought to be the reason for whyeconomics is called the �dismall science�

His key assumption is that technological progress is slow

given the historical record, this was a very reasonableassumption in 1798 (the irony is that the book was publishedjust when technological innovation was about to explode)

While the Malthusian model does poorly in explaining theperiod following its publication, it remains the workhorsemodel for explaining pre-Industrial economic growth



Testing the Malthusian model: The Plague

The Malthusian model suggest that living standards arestagnant in the long-term

in the �short-run� per capita income should �uctuate withpopulation (population ↓ → wages ↑)

The canonial example

the Plague killed between one third and one half of theEuropean population in 1348�50



Real wages in England, 1200�1869Clark (2005, JPE)

working class in england 1311

Fig. 4.—Real wages, 1200–1869, Phelps Brown and Hopkins vs. new series. In bothseries, 1860–69 has been set to 100. Sources: Phelps Brown and Hopkins (1981, 28–31),table A2.

some surprising declines in productivity in between. The seventeenth-century advances in intellectual understanding of the natural world—Francis Bacon, Isaac Newton, Robert Hooke, Robert Boyle, and theirilk—apparently had little effect on the efficiency of the economy before1800.

The series developed here is very different from that of Phelps Brownand Hopkins, however. Figure 4 shows the two series for comparisonfor the decades before 1870. In particular, real wages before 1600 aremuch lower, in some decades being almost 50 percent less than PhelpsBrown and Hopkins’ estimate. The Appendix details why these seriesdiffer so much and why the current estimates are preferable.

The revised series also implies a very different image of economicgrowth in England before the Industrial Revolution. Figure 5 shows realwages by decade with these data from the 1280s to the 1860s versusestimated English population. Now in the decades prior to 1600 thereis a remarkably stable inverse relationship between wages and popula-tion. The curve in figure 5 shows the fitted relationship from regressingthe logarithm of the real wage on the log of population for the decadesof the 1280s to the 1590s. Population alone explains wages very well inthe years before 1640.

With the new data on wages, the efficiency of the economy shows thefirst signs of significantly exceeding medieval levels in the 1640s, when


Note: PBH refers to the earlier estimates by Phelps Brown and Hopkins (1981)



Real wages vs. Population in England, 1280�1869Clark (2005, JPE)

1312 journal of political economy

Fig. 5.—Real wages vs. population on the new series, 1280s–1860s. The line summarizingthe trade-off between population and real wages for the preindustrial era is fitted usingthe data from 1260–69 to 1590–99. Sources: population, same as for fig. 3; real wage,table A2.

real wages are 11 percent higher than would be implied by the popu-lation given the observations before 1600. There was seemingly signif-icant productivity growth in the economy between the 1630s and 1740s.By the 1740s, wages were 67 percent higher than would be predictedfrom the pre-1600 relationship. This growth was followed by an apparentpause in productivity growth at the eve of the classic Industrial Revo-lution, before its resumption in the 1790s. However, real wages in thedecades of the 1770s to the 1810s were depressed by as much as 10percent by the heavy indirect taxes imposed to finance the substantialmilitary expenditures of the government in these years of the AmericanRevolutionary War and the struggle with Napoleon, and by the disrup-tions of trade caused by the wars. The seeming pause in TFP growth inthese years may thus reflect just the limitation of trying to infer TFPgrowth from wage and population information alone.

The beginning of the escape from the Malthusian stagnation in En-gland in the 1640s is a surprise, considering the social and politicalhistory of seventeenth-century England. From the 1630s to the 1680sthere was considerable political and religious conflict, resulting in anopen civil war for most of the 1640s between the king and Parliament.After the execution of King Charles I in 1649, there were 11 years ofunsuccessful rule first by Parliament and then by a military dictatorship


�[...] the break from the Malthusian era of little advance in e�ciency in Englandbegan circa 1640, long before the famous Industrial Revolution, and before even theemergence of the modern political regime in England in 1689�


http://en.wikipedia.org/wiki/The_dismal_science


Testing the Malthusian model: Poor ReliefBoyer (1989, JPE)

Child allowances were commonly paid to laborers with largefamilies in the early 19th century England

Malthus: allowances cause the birth rate to increase

Boyer (1989) argues that Malthus was right

parish level data for 1926�30 (N=214)birth rates positively associated with child allowances(conditional on income, population density, child mortality, housingavailability, presence of allotments of farmland to laborers, presence ofcottage industry, agricultural employment share and distance to London)

Question

what needs to be true for Boyer's estimates to be causal?



Testing the Malthusian model: The MoriscosChaney, Hornbeck (2013)

In 1609, Spain suddenly and unexpectedly expelled roughly300,000 Moriscos (converted Muslims)

In the Kingdom of Valencia, 1/3 of the population was expelled

CH examine the impact of the expulsion to later populationgrowth and agricultural output in Valencia and �nd that it

decreased total output, increased per capita outputincreased the persistence of extractive institutions



BackgroundChaney, Hornbeck (2013)

711: Islamic forces invaded the Iberian Peninsula

many areas prospered economically under the Muslim ruleValencia became one of the largest cities in Western Europe

1238: Valencia took over by the Christians

much of the land and legal jurisdiction ceded to the nobilityMuslims forced to provide labor services, large share of harvestsChristians granted relatively favorable conditions

1525: forced conversion to Christianity

the former-Muslim population became known as Moriscos, who�lived like Christians and payed like Muslims�nobility resisted growing pressure to expel the Moriscos(according to a popular saying: quien tiene moro, tiene oro)



ExpulsionChaney, Hornbeck (2013)

Expulsion ordered in Sept. 1609

approximately 130,000 Moriscos (1/3 of population) exportedMoriscos allowed to take only what they could carry with them

Immediate impact

many nobles bankrupted, creditors faced �nancial ruinValencia's central bank collapsed in 1613

The Crown worked out a �rescue package�

creditors took substantial losses, nobility agreed to pay a sharenobility allowed to impose extractive taxes on Christians



Christian settlementChaney, Hornbeck (2013)

Initially belief that Christians would face the same (low) taxesas elsewhere → migrants began to arrive immediately

Once they realized that taxes would be high, many left

the Crown issued edicts that limited outmigrationtemporary tax concessions to attract migrants ... often in theform of debt that further limited outmigration

1693: (failed) Peasant Revolt

An aside: after the Plague of 1349, England passed quitesimilar Statue of Laborers in 1351, which also led to a (moresuccessful) Peasant Revolt in 1381.

1808: Napoleon declared the abolition of the Spanishseignorial regime



DataChaney, Hornbeck (2013)

Data from tithing auctions

the Archbishopric entitled to about 10% of agricultural outputrather than directly collect agricultural goods, theArchbishopric auctioned the right to collect its sharewinning bid value by district and time are available

Augmented with population data

historical maps: geographic area of each tithing districtpopulation data: town recods for 1569, 1609, 1622, 1646,1692, 1712, 1730, 1768, and 1786


Morisco Population Share in 1609Chaney, Hornbeck (2013)

!!""

!"#$%&'()''*+,-.&'/"01%"2103'*4+5&5'67'89%"029':9-$.+1"9;'*4+%&'";'(<=>'

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

0% Morisco in 31districts (white)

0�100% Morisco in 36districts (light gray)

100% Morisco in 31districts (medium gray)

In the second category,Morisco communitieswere largely segregatedand their averagepopulation share was50%.


Baseline District CharacteristicsChaney, Hornbeck (2013)

Table 1. Baseline District Characteristics, by Morisco Population Share in 1609Pre-ExpulsionSample Mean Basic Difference With Geographic Controls

District Outcome in 1609: (1) (2) (3)Population per km2 9.13 - 0.496** - 0.229

[17.6] (0.181) (0.173)Output per km2 379 - 1.218** - 0.736**

[1256] (0.238) (0.229)Output per capita 30.3 - 0.695** - 0.508**

[19.4] (0.127) (0.125)Notes: Column (1) reports average district characteristics in 1609 (in levels), and the standard deviation is reported in brackets. Output is measured as the auctioned tithe value (in lliures), multiplied by ten. Column (2) reports the basic difference for each district characteristic (in logs) by the district's Morisco population share in 1609: the coefficients are estimated by regressing the indicated county characteristic on the district's Morisco population share in 1609 (between 0 and 1). Column (3) reports the estimated difference when controlling also for seven geographic characteristics of districts (distance to the City of Valencia, distance to the coast, terrain ruggedness, latitude and longitude, average rainfall, and agricultural suitability). There are 98 districts with population data in 1609 and 96 districts with output data in 1609. Robust standard errors are reported in parentheses: ** denotes statistical significance at 1%, * denotes statistical significance at 5%.

Log Difference by 1609 Morisco Share:

!"

Estimates from regressing district-level outcomes in 1609 on districts' Moriscopopulation share in 1609. Column 2 reports that population, output, and output percapita are substantially lower in Morisco districts in 1609. Part of these observeddi�erences can be attributed to observed geographic di�erences between Morisco andChristian districts (distance to the City of Valencia, distance to the coast, averageterrain slope, latitude, longitude, average rainfall, agricultural suitability). Lowermeasured �output� in Morisco districts may partly be a statistical artifact, re�ectingthe historical management of Church tithes.



Preliminary analysisChaney, Hornbeck (2013)

Di�erences-in-di�erences approach

compare changes in districts hardest hit to changes in disrictsthat were less a�ected

Identication assumption

districts with a greater Morisco population share in 1609 wouldhave changed similarly to other districts, if not for theMoriscos' expulsion (conditional on geographic characteristics)


Changes in log populationChaney, Hornbeck (2013)

!"##

!"#$%&'()''*+,"-.,&/'01.2#&+'"2'3"+,%"4,'5$,46-&+7'89':;<='>6%"+46'?6@$A.,"62'B1.%&'!"#$%&'(&&)*+&!*,-%"./*#&

&!"#$%&0(&&)*+&1-.,-.&23/.4$56&

&!"#$%&7(&&)*+&1-.,-.&23/.4$56&,$8&9",/."&

&:*.$5;&&<*8&.4$&/#=/9".$=&*-.9*>$&?"8/"@%$;&&$"94&,"#$%&8$,*8.5&$5./>".$=&94"#+$5&/#&"&A*8>$8%B&CDDE&F*8/59*&=/5.8/9.G&8$%"./?$&.*&94"#+$5&/#&"&A*8>$8%B&DE&F*8/59*&=/5.8/9.(&&H5./>".$5&"8$&A8*>&$I-"./*#&C&/#&.4$&.$J.G&K4/94&9*#.8*%5&A*8&A/J$=&$AA$9.5&@B&=/5.8/9.&"#=&B$"8&"#=&5$?$#&+$*+8",4/9&94"8"9.$8/5./95&/#.$8"9.$=&K/.4&B$"8(&&L"54$=&%/#$5&8$,*8.&MNE&9*#A/=$#9$&/#.$8?"%5('

Estimates for βt from regression lnPdt = βtMoriscod + αt + αd + γtXd + εdt , wherePdt is the population of district d in year t, Moriscod is the population share ofMoriscos in 1609, year �xed-e�ects αt capture changes common to all districts,district �xed e�ects αd caputure time-invariant unobservable di�erences acrossdistricts and Xd is a vector of geographic characteristics. The solid circles indicate thepoint estimates in each year, relative to the omitted base year of 1609, and thevertical lines indicate 95% con�dence intervals.

Changes in log output, output per capitaChaney, Hornbeck (2013)

!"##

!"#$%&'()''*+,"-.,&/'01.2#&+'"2'3"+,%"4,'5$,46-&+7'89':;<='>6%"+46'?6@$A.,"62'B1.%&'!"#$%&'(&&)*+&!*,-%"./*#&

&!"#$%&0(&&)*+&1-.,-.&23/.4$56&

&!"#$%&7(&&)*+&1-.,-.&23/.4$56&,$8&9",/."&

&:*.$5;&&<*8&.4$&/#=/9".$=&*-.9*>$&?"8/"@%$;&&$"94&,"#$%&8$,*8.5&$5./>".$=&94"#+$5&/#&"&A*8>$8%B&CDDE&F*8/59*&=/5.8/9.G&8$%"./?$&.*&94"#+$5&/#&"&A*8>$8%B&DE&F*8/59*&=/5.8/9.(&&H5./>".$5&"8$&A8*>&$I-"./*#&C&/#&.4$&.$J.G&K4/94&9*#.8*%5&A*8&A/J$=&$AA$9.5&@B&=/5.8/9.&"#=&B$"8&"#=&5$?$#&+$*+8",4/9&94"8"9.$8/5./95&/#.$8"9.$=&K/.4&B$"8(&&L"54$=&%/#$5&8$,*8.&MNE&9*#A/=$#9$&/#.$8?"%5('


Preliminary analysis: discussionChaney, Hornbeck (2013)

Former-Morisco districts experienced

relative declines in population and outputrelative increase in output per capita(population remained persistently lower, but output recovered)

Limitations

Malthusian theory implies an explicit dynamic relationship(districts' growth rate depends on their initial outcome value)Morisco and Christian di�erent already prior to the expulsion→ they should experience di�erential growthdi�cult to quantify whether convergence in former-Moriscodistricts was slower than might be expected

PhD students: pay attention to how CH discuss themagnitudes of the estimates (page 17, �rst paragraph)



Estimating convergenceChaney, Hornbeck (2013)

Estimation equation

Yd ,t − Yd ,t−ττ

= δYd ,t−τ + αt + αd + γtXd + βtMoriscod + εdt

where Moriscod is the population share of Moriscos in 1609, year �xed-e�ects αtcapture changes common to all districts, district �xed e�ects αd caputuretime-invariant unobservable di�erences across districts and Xd is a vector ofgeographic characteristics.

Districts modeled as converging to steady-state

annual growth rate depends on the di�erence betweensteady-state and true outcome in period t − τsteady-state unobserved, but allowed to vary by district, yearand geographic characteristics

Econometric challenges: see the paper



ResultsChaney, Hornbeck (2013)

Population growth in former-Morisco areas

substantially lower in 1609�1622...and remained so until 1786note that they should grow faster after loosing population

Output growth in former-Morisco areas

substantially lower in 1609�1622...and remained so for some time (mostly converging along theexpected growth path through the 18th century)

Output per capita in former-Morisco areas

substantially lower in 1609�1622...and remained so until 1786



ConclusionsChaney, Hornbeck (2013)

Extractive institutions

persisted despite increased labor scarcityan example how labor may empower workers, but alsoencourage elites to strengthen e�orts to coerce themslowed population convergence by limiting labor income

Output recovered more quickly than population

an intriguing interpretation is that persistent labor scarcityencouraged labor-saving technology adaptation

Do the results invalidate the Malthusian model?

no, if disposable income did not rise with output per capitaalso: estimates for the entire sample region indicate generallyfast convergence in population and output


Papers for the essays

Young (2005): The Gift of the Dying: The Tragedy of AIDSand the Welfare of Future African Generations, QJE 120:423-466

This paper simulates the impact of the AIDS epidemic using amodel, where the epidemic a�ects (a) human capitalaccumulation and (b) fertility. It argues that the epidemic, onnet, enhances the future per capita consumption possibilities ofthe South African economy.

Voigtländer, Voth (2013): How the West "Invented" FertilityRestriction. AER, 103(6): 2227-64

This paper argues tthat the Plague in 1348�1350 improvedfemale labor market prospects by triggering a shift towards thepastoral sector. As a consequence, the European MarriagePattern (later marriage) emerged and reduced childbirths byapproximately one-third.

Appendix

Appendix Causal questions OLS Dif-in-Dif

The English Clergymen

Malthus was one of the English clergymen, who were bene�ciaries of �asystem that provided [them] with an extremely good living and requiredlittle in return [...] Though no one intended it, the e�ect was to create aclass of well-educated, wealthy people who had immense amounts of timeon their hands. In consequence many of them began, quitespontaneously, to do remarkable things� (Bryson 2010, 34�36)

Other examples include

Edmund Cartwrigth, inventor of the power loomGeorge Bayldon, complier of the �rst dictionary of IcelandicLaurence Sterne, author of popular novelsJack Russell, �inventor� of the Russell terrierWilliam Greenwell, founding father of modern archeaology


Steady States in a Model with Multiple EquilibriaVoigtländer and Voth (2013, REStud)

[14:26 4/4/2013 rds034_online.tex] RESTUD: The Review of Economic Studies Page: 776 774–811

776 REVIEW OF ECONOMIC STUDIES

Wage

Steady state in the standard Malthusian model Steady states with "Horsemen effect"

Dea

th/B

irth

rate birth rate

death rate

C

wC Wage

Dea

th/B

irth

rate

wH

w0

death rate

birth rate

EU

E0

EH

Figure 1

Steady states in the standard Malthusian model and with “Horsemen effect”

it converges to EH . This is equivalent to a “ratchet effect”: Wage gains after the Black Deathbecame permanent.6

The crucial feature to obtain multiple steady states in Figure 1 is an upward-sloping partof the death schedule. This reflects the historical realities of early modern Europe.7 Accordingto Malthus (1826), factors reducing population pressure include “vicious customs with respectto women, great cities, unwholesome manufactures, luxury, pestilence, and war.” We focus onthree—great cities, pestilence, and war. All of them increased in importance after the plaguebecause of higher per capita incomes. High wages were partly spent on manufactured goods,mainly produced in urban areas. Cities in early modern Europe were death-traps, with mortalityfar exceeding fertility rates. Thus, new demand for manufactures pushed up aggregate death rates.War and trade reinforced this effect. Between 1500 and 1800, the continent’s great powers werefighting each other on average for nine years out of every ten (Tilly, 1992). This was deadly mainlybecause armies on the march often spread epidemics. Wars could be financed more easily whenper capita incomes were high. The difference between income and subsistence increased, leavingmore surplus that could be spent on war. In effect, war was a “luxury good” for princes. In addition,trade grew as people became richer, and it also spread germs.8 In this way, the initial rise in wagesafter the Black Death was made permanent by the ‘Horsemen effect,’ pushing up mortality ratesand producing higher per capita incomes. Thus, Europe experienced a simultaneous rise in warfrequency, in deadly disease outbreaks, and in urbanization. The Horsemen of the Apocalypseeffectively acted as “Horsemen of Riches”.9

6. A large positive shock to technology could theoretically also cause this transition. However, pre-modern ratesof productivity growth are much too low to trigger convergence to the high-income steady state.

7. The theoretical and historical conditions under which the “Horsemen effect” led to multiple steady states arediscussed in detail in Section 2.

8. Numerous studies have focused on the interaction between domestic armed conflict and income. Many findthat civil wars decline in frequency after positive growth shocks (Collier and Hoeffler, 1998, 2004; Miguel et al., 2004).In contrast, Grossman (1991) has argued that higher incomes should promote wars (“rapacity” effect), as there is moreto fight over. Martin et al. (2008) find that more multilateral trade can lead to more war.

9. This is the opposite of the negative effect of wars, civil wars, disease, and epidemics on income levels foundin many economies today (cf. Murdoch and Sandler, 2002; Hoeffler and Reynal-Querol, 2003). The main reason for thisdifference is that human capital is crucial for development today, while it was not in pre-modern times, when decreasingreturns to (unskilled) labour in agriculture dominated the production pattern.

by guest on July 29, 2013http://restud.oxfordjournals.org/

Dow

nloaded from

A Malthusian model where death rates increase with income over some range.Steady state E0 combines low income per capita with low mortality, whilesteady state EH is characterized by higher wages and higher mortality. PointEU is an unstable steady state. Suppose that the economy starts out in E0. Amajor positive shock to wages (beyond EU) will trigger a transition to thehigher-income steady state. VV argue that higher income may have increaseddeath rates in Europe by increasing urbanization, trade and con�ict.


Causal questions

We are typically interested in questions suchs as the impact of

increase in living standards on fertility, mortalityculture, institutions etc. on economic growth

These are causal question, i.e. they concern di�erences

between counterfactual states of the world


Potential Outcomes

Easiest to think in the context of a binary "treatment"

potential outcome =

{y1i if Di = 1

y0i if Di = 0

y1i =outcome for unit i if he/she/it receives the �treatment� Di ; y0i =outcome for thesame unit if he/she/it does not receive the treatment

The causal e�ect (for unit i)

[y1i − y0i ]

the di�erence in the potential outcomes with and without the treatment for unit i

But we only observe

yi = yoi + [y1i − y0i ]Di

Identi�cation

The aim is to construct a plausible counterfactual

�reduced form�: a control group that tells us of what wouldhave happened to the treatment group without the treatment�structural models�: estimate parameters of a model that canthen be used to predict counterfactual outcomes

The treatment e�ect is identi�ed if

a valid control group existsa model is a good approximation of reality and a valid controlgroup for identifying its parameters exist

Selection bias

[what would have happened to the treatment group withoutthe treatment] − [what happened to the control group]


Point Estimates and Standard Errors

Most of the time, we are interested in two numbers

Point estimateStandard error

An estimate is statistically signi�cant when it is at least twostandard errors away from zero

An estimate may be statistically insigni�cant because

precisely estimated point estimate is close to zerostandard errors are large (i.e. sample size is too small)→ this does not prove that the true e�ect is zero/small.

Beware of underpowered studies! Read Gelman, Weakliem(2009): Of Beauty, Sex and Power, American Scientist 97


Ordinary Least Squares (OLS)

Estimation equation

yi = αDi + Xiβ + εi

yi = outcome, Di = treatment (0/1 or �exposure�), Xi = observable factors, εi =unobservable factors.

An estimate of α is a

(weighted) average di�erence in y

... between units that have di�erent treatments

... but identical observed characteristics

This is a causal e�ect if

Di and εi are conditionally independent: for units withidentical Xi , treatment is as good as randomly assigned

http://www.stat.usu.edu/jrstevens/biostat/3.5.Beauty.pdf


Di�erences-in-Di�erences (Dif-in-Dif or DD)

Dif-in-dif is based on the assumption that the counterfactualtrend in the treatment and control groups are the same

i.e. there may be permanent unobserved di�erences betweenthe treated and untreated as long as counterfactual changesare paraller

The most basic dif-in-dif setup looks like this

Pre-Period Post-Period Dif

Treatment Group a b b-aControl Group c d d-c

Dif-in-Dif (b-a)-(d-c)where a,b,c and d are means of the outcome


How does Dif-in-Dif work?

Suppose that the true data generating process is

yit = α (Dit × postt) + βpostt + θDi + εit

where yit is the outcome, Dit treatment status (0/1), postt is a dummy for thepost-peridod (they captures general changes between pre- and post-period that a�ectboth the treatment and the control group), Di captures time-invariant di�erencesbetween the treament and control group, and εit are unit speci�c time-variantunobserved factors. That is, the treatment group di�ers from the control group(regarless of the treatment) by the amount θ.

Dif-in-Dif

Period 0: no-one treated. Period 1: treatment group (T=1) treated

E[yT1

]− E

[yT0

]= (α+ β + θ)− (θ)

= α+ β

E[yC1

]− E

[yC0

]= β

Even if θ 6= 0, we have E[(yT1 − yT0

)−

(yC1 − yC0

)]= α

Generalized Dif-in-Dif

Estimation equation

yit = αt (Di × δt) + δt + ηi + εit

yit is the outcome, Di treatment status (0/1 or �exposure�), δt is time-variantunobserved factor that a�ects everyone, and ηi and εit are unit speci�c time-invariantand time-variant unobserved factors. Note that α now has a time subscript t.

Now there may be several pre- and post periods (can estimatethe dynamics of the e�ects)

αt measure the di�erences in changes (in comparison to somebaseline year) between the treatment and control groups

Note that the �main e�ect� of Di is captured by vector ηi , butthe interactions between treatment and year remains identi�ed

Documents

IntroductionCourse logisticsMalthusian …IntroductionCourse logisticsMalthusian modelMoriscosPapers fro the essays Grading and Requirements Essays (50%) wTo essays about paper-pairs